Re: [PATCH 02/39] MIPS: loongson64: Add header files path prefix

[maobibo]

On 03/31/2020 12:21 PM, Jiaxun Yang wrote:
> 
> 
> ä 2020å3æ31æ GMT+08:00 äå11:53:41, maobibo <maobibo@xxxxxxxxxxx> åå:
>>
>>
>> On 03/31/2020 11:34 AM, Jiaxun Yang wrote:
>>>
>>>
>>> --
>>> Jiaxun Yang
>>>
>>>  ---- å ææä, 2020-03-30 18:00:19 maobibo <maobibo@xxxxxxxxxxx> æå
>> ----
>>>  > 
>>>  > 
>>>  > On 03/30/2020 04:55 PM, Jiaxun Yang wrote:
>>>  > > 
>>>  > > 
>>>  > > ä 2020å3æ30æ GMT+08:00 äå4:38:51, bibo mao <maobibo@xxxxxxxxxxx>
>> åå:
>>>  > >> Remove header files from arch/mips/include/asm/mach-loongson64
>>>  > >> to arch/mips/loongson64/include/mach
>>>  > >>
>>>  > >> Signed-off-by: bibo mao <maobibo@xxxxxxxxxxx>
>>>  > > 
>>>  > > Hi Bibo,
>>>  > > 
>>>  > > Thanks for your patch.
>>>  > > 
>>>  > > What's your intention?
>>>  > > Did you meet any problem about headers?
>>>  > > 
>>>  > 
>>>  > Hi Jiaxun,
>>>  > 
>>>  > When I add irqchip support for mips kvm, file virt/kvm/irqchip.c
>>>  > requires local header file arch/mips/kvm/irq.h, it fails to
>> compile.
>>>  > Since there is file with the same name located at:
>>>  >   arch/mips/include/asm/mach-generic/irq.h
>>>  >   arch/mips/include/asm/mach-loongson64/irq.h
>>>  > 
>>>  > The compiler fails to choose the correct header file irq.h,
>>>
>>> You'd better rename kvm's irq.h in this case.
>> why should we rename kvm's irq.h?  It works on other arches like
>> x86/aarch/powerpc, it does not work on mips platform. Just because
>> these
>> header files are not used with path prefix on mips system, it brings
>> out
>> such problem.
>>
>> Today if we modify generic kvm code, it maybe brings out problem with
>> other modules tommorrow. Else we declare that these files can be
>> defined
>> locally on all modules:
>>  cpu-feature-overrides.h  floppy.h  ide.h  ioremap.h  irq.h
>>  kernel-entry-init.h  kmalloc.h  mangle-port.h  mc146818rtc.h
> 
> It sounds reasonable.
> But I just don't want to rush into this kind of tree wide
> "scrub" unless the problem is urgent.
> 
> For your problem, a simple workaround might be fine.
> 
> And there might be some reference to these mach headers in other subsystems.
> 
> Please examine them extremely carefully and do compile test with
> all yes config for all the platforms.
> 
> Thomas said he is going to drop some platform support in this cycle,
> so you can do less work after that.

I also think that it is a big patch for just one conflict header file,
however it is actually one issue on mips platform.

This patch set is not urgent, it will be better if Thomas drop some
platforms support. It's just my option to move headers to platform
specific directory, We can discuss for better method, after all it is
common code. And I actually do compile test with all defconfig.

If we need test with all yes config for all the platforms, does anyone
do so if common code is changed? Maybe he can teach me how to do it.

regards
bibo,mao

> 
> I'll miss these header hacks anyway.
> 
> Thanks.
> 
>>
>> regard
>> bibo,mao
>>
>>>
>>>  > 
>>>  > regards
>>>  > bibo,mao
>>>  >